Summary

We covered a lot of ground in this chapter. Network management problems can be solved . Creating NMS that incorporate standard GUI paradigms is very important for increasing usability and reducing end- user training time. Providing NMS that incorporate well-engineered solutions is a useful guide to developers. Components form a useful abstraction for building durable management features. Components should provide ease of use for developers, and relationships between them should match those found in real-world objects, such as Frame Relay interworking across ATM, virtual connections, and EROs. The broad range of expertise in NMS vendor organizations can be leveraged to produce high-grade solutions. NMS can themselves be improved to support increasingly popular features like end-to-end services, particularly when standards like PWE3 emerge (an important point about PWE3 is that it is edge-to-edge but enables end-to-end services and QoS).

The stovepipe structure of many NMS was described along with some consequences of the fact that FCAPS applications usually share a database and a host machine. MIBs can also stand some improvements in the shape of default values, loose column coupling, and centralized features. The merits of a single data model were enumerated, with an unexpected benefit being thinner (simpler) provisioning code. Less code helps in speeding up development.

Distributed NMS applications provide some advantages in the form of less heavily loaded hosts , but this may be at the expense of more network traffic. Raising the intelligence of network devices is another good way of improving overall network function and manageability. This involves getting NEs to do more work, including preprocessing (or aggregating) traps and supporting policy-based management.

PBNM was introduced as an emerging model for management. By pushing policies onto NEs, the philosophy of PBNM is to treat the network as a type of computer. This allows for a more fine-grained management model than is available using a connection- and node-based approach. The QoS mechanisms in the IP and MPLS networks were discussed. Policies were described prior to introducing COPS and COPS-PR for policy distribution between clients and servers.

An increasingly advanced range of network processors is helping to encourage adoption of PBNM by allowing advanced traffic management functions. PBNM can be leveraged to the greatest extent when it is deployed in conjunction with DEN. This is a powerful means of modeling and controlling the managed network.

Integrated and Differentiated Services were described in the context of IP and MPLS. Layer 3 QoS is efficiently enabled using these technologies. Attacks on networks continue to pose a substantial risk, and this is perhaps even more serious with DiffServ networks.